Comp. by:ananthi Date:24/10/05 Time:12:59:03 Stage:First Proof File Path:// spsind002s/cup_prod1/PRODENV/000000~1/00DA26~1/S00000~2/00D226~1/ 000000~4/000007342.3D Proof by: QC by: 12 Modelling of large herbivore – vegetation interactions in a landscape context PETER J. WEISBERG, MICHAEL B.COUGHENOUR AND HARALD BUGMANN INTRODUCTION There is growing appreciation of the important role large herbivores can play in vegetation, ecosystem and landscape dynamics (Hobbs 1996, Danell et al. 2003, Rooney & Waller 2003, and earlier chapters of this volume). In turn, there has been an improved understanding of the importance of landscape pattern for large herbivore dynamics (Turner et al. 1994, Illius & O’Connor 2000, Walters 2001), and research into patterns of animal movement through landscapes (Gross et al. 1995, Schaefer et al. 2000, Johnson et al. 2002). At landscape scales, the large herbivore-vegetation interaction can be quite complex, involving many interacting factors such as plant competition, landscape pattern, climate, disturbance regimes and biogeochemical cycles. The earlier chapters of this volume demonstrate the complexity of such relationships, and the difficulty in establishing simple generalizations. Simulation modelling has proved a useful tool for disentangling some of this complexity, and for integrating information across multiple scales. There are numerous modelling approaches, at varying levels of complexity, developed to satisfy different research objectives, for simulating the impacts of large herbivores upon vegetation or vice versa. However, few represent key interactions between the two ecosystem components in a balanced manner. In this chapter, we review the different modelling approaches for representing large herbivore-landscape interactions in an integrated way. Large Herbivore Ecology, Ecosystem Dynamics and Conservation, ed. K. Danell, P. Duncan, R. Bergstro¨m & J. Pastor. Published by Cambridge University Press. # Cambridge University Press 2006. Comp. by:ananthi Date:24/10/05 Time:12:59:03 Stage:First Proof File Path:// spsind002s/cup_prod1/PRODENV/000000~1/00DA26~1/S00000~2/00D226~1/ 000000~4/000007342.3D Proof by: QC by: Landscape modelling of herbivores and vegetation 349 By integrated models, we refer to modelling approaches that consider vegetation and animal dynamics with similar levels of complexity, bridging the two key components through the ecological process of herbivory. Integrated grazing models have been used to address a number of eco- logical questions that consider sufficiently long time scales for feedbacks between large herbivores and vegetation to become important. Due to our emphasis on the landscape context, we will focus on spatially explicit models. The major challenges inherent in such modelling approaches are discussed, particularly problems related to scale and constructing models of use for management and conservation. It is not our goal to provide solutions to all of these difficulties, but rather to synthesize the scope of the problem, and to briefly summarize how these challenges are or are not addressed by the current generation of integrated large herbivore-vegetation models. We hope that by describing the current limitations for such models, we identify critical gaps in our knowledge of how large herbivores and vegetation interact in complex landscape systems. MODELLING APPROACHES Three general approaches to modelling large herbivore-vegetation pro- cesses can be characterized: animal-focused, plant-focused, and integrated (Fig. 12.1). We briefly discuss the former two approaches, focusing the remainder of our chapter on the latter. Approaches focusing on large herbivore dynamics When the questions of interest focus on animal physiology or population dynamics, vegetation may be portrayed as a single input variable for available forage or, if foraging ecology is irrelevant for the model, available energy (Fig. 12.1a). Forage or energy intake is balanced against the large herbivore’s energetic requirements (basal metabolism, thermal metabol- ism, requirements for travel, foraging and lactation), often using a simple ‘input-output’ energy budget approach. The energetic state of the animal typically influences reproductive success through population parameters determining mortality and fecundity. Populations may then be distributed over some larger area either through explicit simulation of their movement patterns, or using a ‘fly and sit’ approach where habitat selection functions are related to landscape pattern as represented by a spatial database. Distribution determines the local population density, which in turn influences future intake rates for each simulated patch. Comp. by:ananthi Date:24/10/05 Time:12:59:03 Stage:First Proof File Path:// spsind002s/cup_prod1/PRODENV/000000~1/00DA26~1/S00000~2/00D226~1/ 000000~4/000007342.3D Proof by: QC by: 350 Peter J. Weisberg, Michael B. Coughenour and Harald Bugmann Figure 12.1. Three general approaches to modeling interactions among large herbivores and vegetation: (a) animal-focused, plant-focused, and (c) integrated. Such animal-focused models will generally require two inputs that relate to vegetation: (1) the amount and quality of available forage, and (2) the mosaic pattern and availability of suitable habitat. In some cases, the models use the spatial pattern of forage availability itself as the habitat map (e.g. Moen et al. 1997). Plant production, which varies seasonally and with the weather, represents the critical driving variable, but there is little or no feedback of herbivory to plant productivity. A single forage quantity representing vegetation at each patch may be decremented due to brows- ing and decomposition, but plant responses to browsing (e.g. compen- satory growth, structural and compositional changes) are not represented. By varying the forage input seasonally, annually or according to scenarios of vegetation change, such a model can be used to describe the influence of variation in forage resources on large herbivore dynamics (Turner et al. Comp. by:ananthi Date:24/10/05 Time:12:59:04 Stage:First Proof File Path:// spsind002s/cup_prod1/PRODENV/000000~1/00DA26~1/S00000~2/00D226~1/ 000000~4/000007342.3D Proof by: QC by: Landscape modelling of herbivores and vegetation 351 1994, Illius & O’Connor 2000). Similarly, animal distribution or move- ment patterns can be made to vary with landscape patterns that vary over time (e.g. Loza et al. 1992). However, such models do not describe the influences of large herbivore effects on plant production or landscape pattern, and so do not represent potentially important feedbacks between plant and animal processes (e.g. resource depletion, stimulation of com- pensatory growth, nutrient relocation, competitive influences, trophic cascades). Therefore, although models of the type shown in Fig. 12.1a may prove theoretically interesting, they are unlikely to provide realistic results for actual landscapes over time periods sufficiently long for plant- animal feedbacks to become important. Regardless, most carrying cap- acity models, foraging models and animal population models fall into the strictly animal-focused category (e.g. Hobbs & Swift 1985, Loza et al. 1992, Xie et al. 1999). We advocate the utility of more integrated model- ling approaches for estimating carrying capacity and appropriate large herbivore population objectives in the context of spatio-temporal variation and long time periods (e.g. Weisberg et al. 2002). Approaches focusing on vegetation dynamics Another approach to modelling large herbivore-vegetation processes em- phasizes the vegetation side (Fig. 12.1b). Where the focus has been on production or population dynamics of vegetation, herbivory has most often been represented simply as the removal of biomass or individual plants. Plant-focused models of herbivory generally take one of two forms: (1) biomass-based production models with grazing as a driving variable, and (2) forest gap models with a browsing effect on tree regeneration. In neither case are feedbacks of altered vegetation composition, spatial pattern or forage availability upon the herbivores explicitly incorporated. Biomass-based models of plant productivity are usually applied to grasslands and other vegetation types dominated by grazers and herb- aceous vegetation, where population dynamics at the level of individual plants are neither important nor feasible to consider. These types of models can represent the effects of herbivory without modelling actual animals associating reductions in plant biomass with particular grazing intensities (Fig. 12.1b). Forage biomass may influence simulated grazing intensity, where large herbivore intake rate is reduced at low biomass levels (representing inefficiency of foraging in resource-poor patches, or animal decisions to forage in more resource-rich patches). However, there is no long-term feedback on grazing intensity due to reductions in large herbivore condition or population levels, or shifts in spatial allocation of foraging Comp. by:ananthi Date:24/10/05 Time:12:59:04 Stage:First Proof File Path:// spsind002s/cup_prod1/PRODENV/000000~1/00DA26~1/S00000~2/00D226~1/ 000000~4/000007342.3D Proof by: QC by: 352 Peter J. Weisberg, Michael B. Coughenour and Harald Bugmann effort, as a result of resource depletion. Jeltsch et al. (1997) used such a model to explore the relative influences of cattle grazing and precipitation on shrub encroachment processes in the South African savanna. In their model, production is determined empirically from the amount